The direct RA-target Hnf1b is required for pancreas development

We identified the transcription factor Hnf1b as direct RA-target gene in the dorsal endoderm during gastrulation and could demonstrate that its function is required for pancreas development in Xenopus. In mouse, Hnf1b was found to be expressed in the endodermal germ layer and in endodermal derived structures including pancreas, liver, gallbladder and duodenum (Barbacci et al., 1999; Haumaitre et al., 2003). We could detect Hnf1b transcripts in the entire endoderm during gastrulation with a minor enrichment in the dorsal endoderm. At later stages, Hnf1b expression was observed in the foregut endoderm including the anterior archenteron roof that gives rise to the dorsal pancreatic anlagen. These observations are consistent with findings in mouse where Hnf1b expression was also detected in early proliferating pancreatic progenitor epithelium together with Pdx1 and Ptf1a (Haumaitre et al., 2005). Furthermore, Hnf1b was identified as pancreatic trunk marker in mouse and lineage tracing experiments showed that Hnf1b-positive cells of the pancreatic epithelium are precursors of acinar, duct and endocrine cells (Solar et al., 2009).

We could further show that the dorsal endodermal expression of Hnf1b during gastrulation is RA-responsive. So far, the induction of Hnf1b by RA was solely described for hindbrain tissue (reviewed in Glover et al., 2006) and the only evidence for endodermal RA-inducibility was found in gastric organoids derived from mouse embryonal stem cells (McCracken et al., 2014). We further identified Hnf1b as direct RA-target what is supported by the identification of two RAREs in the promotor sequence and within the fourth intron of the mouse Hnf1b gene locus (Power and Cereghini, 1995; Pouilhe et al., 2007). The RA-responsive expression in the dorsal endoderm strongly indicates a role for Hnf1b in the pre-arrangement of conditions promoting pancreatic fate. Indeed, upon the downregulation of functional Hnf1b in pancreatic organoids almost a complete loss of pancreatic markers was observed. Furthermore, the endodermal marker Darmin was also decreased whereas another endodermal marker Sox17a was only slightly affected. Darmin is described in Xenopus as an endodermal marker with unknown function (Pera et al.,

101 2003) and no indications for a regulation of Darmin by Hnf1b are found in the literature. Interestingly, Hnf1b and Darmin exhibit a similar expression in the entire endoderm in Xenopus blastula embryos and we find both genes to be RA-inducible expressed in programed explants. As we found Darmin downstream of Hnf1b, Darmin could be identified as novel Hnf1b target genes with a possible role in pancreas development. The rescue of most pancreatic markers and Darmin by a hormone-inducible Hnf1b indicates that the observed knockdown-phenotype is specific due to the downregulation of functional Hnf1b. The data observed for the rescue of Ptf1a expression were inconsistent. Some rescue experiments showed a Ptf1a expression (not shown) and others not. The failed rescue of Ptf1a does not reflect a failed rescue of exocrine differentiation as Pdia2 expression could be rescued. We have no convincing explanations for this observation, but we can provide some speculations. Hnf1b and Ptf1a are found to be co-expressed in pancreatic progenitors and later during the segregation of tip and trunk domains Ptf1a is expressed in acinar cells whereas Hnf1b expression is observed in bi-potential trunk cells where it promotes endocrine differentiation (reviewed in Pan and Wright, 2011; De Vas et al., 2015). The Hnf1b-GR fusion protein, used for rescue approaches, is probably more stable than endogenous protein and continuous DEX-treatment provides high levels of active Hnf1b. It is possible that high Hnf1b levels promote endocrine fate at the expense of exocrine fate. This would be supported by the low transcript levels of Pdia2 in the rescue. However, we cannot provide evidences for this.

Nevertheless, our data indicate a function of Hnf1b in pancreas development and support and confirm previous studies. Mutations within the Hnf1b gene locus in humans cause a monogenetic form of diabetes, named MODY 5 (maturity onset diabetes of the young) and is further linked to kidney and genital malformations (reviewed in Ryffel, 2001; Wang et al., 2004; Haumaitre et al., 2006). Chimeric mutant mice with a specific loss of Hnf1b in the visceral endoderm exhibit a lack of the ventral pancreas and a hypoblastic dorsal pancreas (Haumaitre et al., 2005).

Therefore, Hnf1b is considered as putative upstream regulator of pancreatic progenitor markers Pdx1 and Ptf1a. Our findings in pancreatic organoids could be confirmed in whole embryos where the downregulation of Hnf1b leads to a decreased expression of Pdx1, Ptf1a and Insulin. Moreover, upon Hnf1b over-expression, endodermal Ptf1a and Pdx1 expression domains were significantly expanded. This indicates an early function of Hnf1b in promoting pancreas progenitor formation. Further indications for an early function of Hnf1b in pancreas development were previously found by protein-DNA interaction analysis in mouse,

102

where Hnf1b was found to directly induce Hnf6 expression (Poll et al., 2006). Hnf6 precedes the expression of Pdx1 in the foregut-midgut region and is later restricted to the liver and the pancreas (Landry et al., 1997; Rausa et al., 1997). Hnf6 was shown to initiate Pdx1 expression in mouse and the inactivation of Hnf6 results in a delayed onset of Pdx1 expression leading to pancreatic hypoplasia (Jacquemin et al., 2003). Therefore, it is suggested that a sequential transcriptional cascade of Hnf1b, Hnf6 and Pdx1 directs endodermal cells into pancreatic progenitors (Poll et al., 2006). However, the fact that the Pdx1 expression in Hnf6 mutants is delayed and not completely missing indicates the requirement of additional factors, possibly further Hnf1b targets.

Besides the indicated early function of Hnf1b in progenitor formation, De Vas and colleagues demonstrated in a mouse system the requirement of Hnf1b for endocrine cell specification (De Vas et al., 2015). They inactivated Hnf1b specifically in pancreatic progenitors and observed the absence of Ngn3-positive endocrine precursor cells throughout embryogenesis. Another direct Hnf1b-target Hnf4a is also linked to MODY as β-cells in patients with Hnf4a mutations exhibit an impaired insulin secretory response to glucose (Yamagata et al., 1996; Hattersley, 1998;

Lausen et al., 2000). Pancreatic organoids would be a suitable system for the identification of further direct Hnf1b targets that are possibly involved in pancreas development.

In summary, our data confirm several studies that found Hnf1b as direct RA-target and its requirement for pancreas development. We could provide further evidences for an early endodermal induction of Hnf1b by RA and its early function in pancreatic progenitor formation. However, the gene network induced by Hnf1b that is involved in pancreas development needs to be further investigated. Moreover, we show that Hnf1b is not sufficient to substitute for RA in pancreas development. Hence, Hnf1b is not the only RA-responsive gene that is required for pancreas development.

103 4.4 The direct RA-target Fzd4 is required for pancreas development

We identified the transmembrane Wnt-receptor Fzd4 and its secreted splice variant Fzd4s as direct RA-target genes required for pancreas development. Fzd4 was previously found to be upregulated together with other receptors and Wnt-ligands by RA in human embryonal tumor cells (Katoh, 2002). Furthermore, evidences exist for an interplay of RA-signaling and Wnt-signaling in chondrocytes to regulate cartilage matrix homeostasis (Yasuhara et al., 2010). However, a connection of Fzd4 and RA-signaling in pancreas development is not described so far.

Two studies of Fzd4 in Xenopus describe an expression in the prospective neuro-ectoderm and later in the head region including the forebrain (Shi and Boucaut, 2000; Zhang et al., 2011). We found an additional Fzd4 RA-responsive expression domain in the dorsal endoderm of gastrula stage embryos. We initially observed this endodermal Fzd4 expression domain in RA-treated embryos only. A prolonged staining of un-treated embryos confirmed a weak endogenous endodermal expression of Fzd4 that have been missed by a too short staining. At later stages, Fzd4 expression was also detected in the foregut. For human and Xenopus Fzd4, an alternative splice variant through intron retention was identified (Sagara et al., 2001; Swain et al., 2005). This Fzd4 variant, named Fzd4s, encodes a small protein missing the transmembrane domain but still containing the Wnt-ligand binding cysteine rich domain (Swain et al., 2005). Therefore, the question appeared which variant is expressed in the dorsal endoderm. The endodermal expression domain was detected by the use of an antisense probe targeting the exonic sequence. This should theoretically detect both variants. Therefore, we additionally used an antisense probe targeting a part of the intronic region that should specifically detect Fzd4s. Steinbeisser and colleagues described an ubiquitous staining of Fzd4s transcripts during gastrulation and at tailbud stages in head region, not identical with Fzd4 but with overlaps in the eye (Swain et al., 2005). In contrast, beside a certain level of staining in the whole embryo, we could detect enriched regions in the mesoderm during gastrulation and later in the notochord and head region. The expression in the notochord was not observed with the probe that targets the exonic regions. Either the probe that targets the exons detects only Fzd4 and not Fzd4s or the probe targeting the intronic region results in an unspecific signal. Thus, we cannot conclude with certainty which variant exhibits the dorsal endodermal expression domain.

104

RNA-sequencing and RT-PCR data indicate that both variants of Fzd4 were induced by RA. However, compared to the other direct RA-target genes Hnf1b and Cyp26a1, the Fzd4 variants exhibit only a minor induction by RA. A certain level of Fzd4 and Fzd4s transcripts is detected also in the absence of RA. This probably corresponds to maternally provided transcripts in the case of Fzd4 (Shi and Boucaut, 2000) and for Fzd4s an additional RA-independent expression is conceivable.

The described loss-of-function studies through morpholino-oligonucleotide and CRISPR/Cas-system should affect both Fzd4 variants. Therefore, Fzd4 and Fzd4s were designated as Fzd4/Fzd4s in this context. We demonstrated that both, the knockdown of Fzd4/Fzd4s by morpholino-oligonucleotide and the knockout by RNA-guided Cas9 in pancreatic organoids showed a similar decrease of pancreatic marker gene expression. For the knockdown approach, several strategies failed to rescue the observed phenotype in order to demonstrate the specifity of the used morpholino-oligonucleotide. As mentioned in the results section, this might have several reasons. On the one hand, we were not able to provide active Fzd4 or Fzd4s specifically during gastrulation. On the other hand, we are not aware of the variant that is required and the appropriate level of active protein necessary for a proper pancreas development.

Both variants were not described to be involved in pancreas development so far.

The transmembrane Fzd4 is described as mediator of both, canonical and non-canonical Wnt-signaling downstream of Wnt5a (Umbhauer et al., 2000; Chen et al., 2003; Mikels and Nusse, 2006; Xu et al., 2004). The activation of non-canonical/PCP pathway by Wnt5a/Fzd4-interaction was shown to be required for arterial network formation in mouse (Descamps et al., 2012). Furthermore, Fzd4 is the only known receptor that binds Norrin and thereby activates canonical Wnt-signaling promoting retinal vascularization in mouse (Xu et al., 2004; reviewed in Ye et al., 2010). Fzd4s shows structural similarities to secreted frizzled-related proteins (sfrps) and is therefore considered to be secreted (Rattner et al., 1997). Fzd4s still contains the Wnt-ligand binding domain and was described in Xenopus as an activator or inhibitor of canonical Wnt-signaling dependent on its ligand. Steinbeisser and colleagues showed that dependent on the corresponding Wnt-ligand, the Wnt/Fzd4s complex is recognized by the LRP5/6-co-receptor and mediates Wnt-signaling or it is not recognized (Swain et al., 2005). Following studies revealed that the activatory or inhibitory function of Fzd4s is further dependent on its concentration. Low Fzd4s concentrations enhance and high concentrations repress Wnt-signaling in the presence of low Wnt-ligand levels (Gorny et al., 2013). This supports the assumption of the importance of appropriate Fzd4/Fzd4s levels

105 required to rescue the morpholino-mediated knockdown phenotype. Findings by the Fzd4/Fzd4s-knockout confirm the results of Fzd4/Fzd4s-knockdown in pancreatic organoids. We could demonstrate that the observed decrease of pancreatic marker gene expression is due to specific mutations within the Fzd4 gene locus as tested potential off-targets showed no mutations.

The next important question that needs to be answered is whether the function of Fzd4/Fzd4s in Wnt-signaling is required for pancreas development. For Fzd4 also a signaling function outside of the classical Wnt-signaling pathway was described.

During Wnt-signaling, Fzd4 usually binds intracellular Disheveled via a PDZ-binding motif that mediates Wnt-signaling (Wong et al., 2003; reviewed in Niehrs, 2012). In a mouse neuronal cell culture, the intracellular PDZ-binding motif of Fzd4 was shown to interact with other PDZ-domain containing proteins than Dvl and thereby promoting dendrite outgrowth (Bian et al., 2015). Therefore, a function of Fzd4 in pancreas development independent from the Wnt-signaling pathway is possible.

However, results from luciferase reporter assays showed an inhibitory effect of RA-treatment on canonical as well as non-canonical reporter activity in Vegt/Noggin-programed explants. The activatory effect of the Fzd4-knockdown on non-canonical reporter activity in these explants suggests the involvement of Wnt-signaling in early pancreas development. This issue will be discussed in the next section. In summary, we identified Fzd4 and its splice variant Fzd4s as direct RA-target genes. Our results strongly suggest the requirement of Fzd4 and/or Fzd4s in pancreas development.

4.5 Wnt- and RA-signaling in pancreas development

We identified the Wnt-receptor Fzd4 and its secreted variant Fzd4s as direct RA-targets with a possible role in pancreas development. Therefore, we asked about the role of Wnt-signaling in pancreas development. Previous studies proposed that Wnt-signaling needs to be repressed for foregut maintenance and therefore to allow a proper pancreas development. In Xenopus, McLin and colleagues found several foregut markers repressed upon Wnt8 over-expression (McLin etal., 2007).

Furthermore, Li and colleagues identified the secreted Wnt-inhibitor sfrp5 to be expressed in the early foregut epithelium of Xenopus embryos. A downregulation of functional sfrp5 leads to smaller foregut domains and in contrast the ectopic sfrp5 expression results in an expanded foregut domain at the expense of the hindgut (Li et al., 2008). However, a more recent study found evidences for the requirement of

106

low Wnt-signaling activity for foregut maintenance. It was shown that the depletion of Wnt-receptor Fzd7 in the foregut results in pancreas agenesis in Xenopus and that a low expression level of this Wnt-receptor is essential for foregut maintenance (Zhang et al., 2013a). Moreover, sfrps that were initially thought to be exclusively negative Wnt-signaling modulators emerged as biphasic regulators in a concentration dependent manner (Mii and Taira, 2009). These findings suggest a regulatory mechanism that ensures an appropriate Wnt-signaling activity in the foregut.

Our finding of Wnt-receptors Fzd4 and Fzd4s involved in pancreas development is further supported by transcriptome analysis of hepatic and pancreatic progenitors in mouse. The transcriptome of bi-potential hepato-pancreatic progenitors was compared to the transcriptome of developed dorsal and ventral pancreatic buds and the liver bud (Rodríguez-Seguel et al., 2013). They found intracellular Wnt-signaling transducers like Disheveled to be expressed equally in all samples whereas Wnt-ligands, receptors and co-receptors were strongly downregulated in liver progenitors. Among these differentially expressed Wnt-components, Fzd4 and its ligand Wnt5a were found. It was demonstrated that endodermal explants from Xenopus embryos treated with soluble Wnt5a exhibit an enhanced expression of Pdx1 and Ptf1a. Furthermore, liver cells treated with Wnt5a strongly induce Pdx1 expression (Rodríguez-Seguel et al., 2013). They suggest that non-canonical Wnt-signaling is a potential promotor of pancreatic fate. We found only Fzd4/Fzd4s expression regulated by RA and not Wnt5a expression. However, we observed an effect of treatment on Wnt-signaling in our explant system. Two hours after RA-addition the activity of both, canonical and non-canonical Wnt-signaling reporter was decreased. Thereby, the non-canonical Wnt-reporter was slightly stronger affected.

The described effect of RA-treatment on canonical Wnt-signaling is consistent with several other studies. Zhang and colleagues identified Ndrg1a as RA-target and demonstrated that the inhibitory function of Ndrg1 on canonical Wnt-signaling is required for foregut development (Zhang et al., 2013b). However, Ndrg1 was not differentially expressed upon RA-treatment in our system. One explanation for this could be the late time point of Ndrg1 induction by RA that was observed in stage 16 embryos earliest, but we searched for RA-targets that were induced within two hours after RA-addition. Another study using mouse ESCs also found the negative regulatory effect of RA on canonical Wnt-signaling (Osei-sarfo and Gudas, 2014). In addition, they found non-canonical Wnt-signaling activated by RA. This activatory effect of RA on non-canonical Wnt-signaling was also described by Harada and colleagues (Harada et al., 2007). There, RA-inducible G-protein-coupled receptors

107 were found to bind receptors and thereby activating non-canonical Wnt-signaling. These findings seem to be contradictory to our observed negative regulation of non-canonical Wnt-reporter activity by RA-treatment. However, the term “non-canonical Wnt-signaling” comprises two different pathways. The planar cell polarity (PCP) pathway that involves Rho GTPase and JNK and on the other hand the calcium pathway that involves calcium-sensitive kinases and PKC (reviewed in Nusse, 2012). Our non-canonical Wnt-reporter system is based on an Atf2-response element that is activated by the PCP-pathway (Ohkawara and Niehrs, 2010). In contrast, both studies used a reporter that contains a binding site for the transcription factor NFAT. This transcription factor is activated by the Wnt/calcium pathway (Dejmek et al., 2006). Thus, the effect of RA on the NFAT-reporter needs to be tested in our system. It is possible that RA-signaling has a biphasic activity on different non-canonical Wnt-signaling pathways.

We further examined the effect of Fzd4-knockdown on Wnt-signaling reporter activity in the explant system. We found canonical Wnt-reporter activity only slightly and not significantly increased upon Fzd4-downregulation. In contrast, non-canonical Atf2-reporter activity was significantly increased. This finding complies with the observed decrease in non-canonical Wnt-reporter activity upon RA addition.

Hence, these data suggest that the negative regulatory effect of RA on non-canonical Wnt-signaling is mediated by Fzd4 and/or Fzd4s. However, Fzd4 as well as Fzd4s were shown to positively regulate non-canonical Wnt/PCP-signaling (Descamps et al., 2012; Gorny et al., 2013). Therefore, the effect of Fzd4-downregulation on non-canonical Wnt-signaling needs to be further investigated.

Moreover, it remains to be tested whether the downregulation of Wnt-signaling by RA is mediated by Fzd4 and if this Fzd4-function is required for pancreas specification.

A connection of Hnf1b and Wnt-signaling was shown in zebrafish. Lancman and colleagues demonstrated that Hnf1b and Wnt2b synergistically function in the specification of hepato-pancreatic progenitors (Lancman et al., 2013). Thus, it is necessary to examine if a combined activity of Hnf1b and Fzd4 and/or Fzd4s is sufficient to substitute for RA in pancreas specification.

108

4.6 Conclusions

In this study, a system of in vitro generated pancreatic organoids was used for the identification of RA-target genes involved in the early pancreas development. We could identify 22 RA-responsive genes, some of which have previously been described as RA-targets. For the transcription factor Hnf1b and the Wnt-receptor

In this study, a system of in vitro generated pancreatic organoids was used for the identification of RA-target genes involved in the early pancreas development. We could identify 22 RA-responsive genes, some of which have previously been described as RA-targets. For the transcription factor Hnf1b and the Wnt-receptor